The presense of cancer within the liver carries an ominous prognosis. The recent development of a totally implanted drug delivery system for hepatic arterial chemotherapy may provide a method to more effectively treat hepatic cancer. This system makes possible reliable long-term hepatic arterial infusion to the entire tumor-bearing liver on a convenient outpatient basis. A major objective of this proposal will be to compare the response rate and toxicity, in a randomized prospective study, of hepatic arterial fluorodeoxyuridine (FUDR) versus systemic intravenous fluorouracil (FU) in patients with colorectal cancer confined to the liver. Patients having no evidence for extrahepatic disease (other than a resectable primary) by non-invasive studies will have a laparotomy for staging and for implantation of a hepatic artery catheter and subcutaneous Model 400 Infusaid pump (Infusaid Corp., Sharon, MA). Thereafter, patients free of extrahepatic disease with the drug delivery system successfully placed will be randomized to receive initially either intravenous FU or hepatic arterial FUDR. Patients will be evaluated for response at 2 month intervals using liver scans and/or computed tomographic xray scans. Patients failing FU will be crossed over to hepatic arterial FUDR. Patients failing FUDR will be candidates for a mitomycin/microsphere study. Such patients, plus those found to have extrahepatic disease at laporatomy, will be entered into randomized study examining whether biodegradable starch microspheres (40 Mum dia), which can potentially deliver more drug selectively to the liver and liver tumor, affect the toxicity and response rate of hepatic arterial mitomycin. Patients who cannot tolerate a test dose of starch microspheres will enter a phase II study of hepatic arterial dichloromethotrexate, an agent which has shown some degree of activity in preliminary studies. New phase II regimens with appropriate pharmacologic rational for hepatic arterial infusion will also be tested. This proposal is similar to those of other members of a potential hepatic arterial treatment consortium. The consortium should be able to perform studies reliably and efficiently due to combined expertise and an enlarged patient population.